Assigning a unique ID to each request in Django

It is a standard practice to assign a unique ID to every request, so that you can easily associate the log messages with their request and, thus, make debugging easier.

You can do that in Django using a middleware. The middleware needs to define the process_request metho and generate the unique ID (e.g. using UUID) to associate with the request.

One way to use this ID is to pass it as an argument down the call stack and use it in every log message. That approach, however, creates a dependency from the upper layer (HTTP layer) to the lower level layers (model or domain layer). Moreover, this approach requires that every function accepts the ID as an argument, which complicates the function signatures.

An alternative is to use thread local variables. These variables are similar to global variables, but their scope is limited within the thread of execution. This way all logging code has access to the thread local variable (if defined).

It should be noted, however, that these are global variables and so should be handled with care.

It is useful, however, to use the same ID in celery tasks, so that you can trace the request in the celery workers log messages, as well. This can be achieved using celery’s before_task_publish_handler signal (to add the ID to the task arguments) and task_prerun_handler signal (to read the ID from the task arguments and set the thread local variable).

The final piece of the puzzle is creating log filters that read the thread local variables and make them available to the log message formatters automatically, e.g.

import logging
from threading import local

_locals = local()


def get_current_identifier():
    """Get the current identifier of the request/task etc."""
    return getattr(_locals, 'identifier', u'')


class ContextFilter(logging.Filter):
    """Add the request.user to the context of a log record."""

    def filter(self, record):
        record.user_identifier = get_current_identifier()
        return True